Task-Oriented Intelligent Solution to Measure Parkinson's Disease Tremor Severity.

Area of Science:

• Neurology
• Biomedical Engineering
• Data Science

Background:

• Tremor is a primary symptom of Parkinson's disease (PD), impacting patient quality of life.
• Current clinical assessment using the MDS-UPDRS Rating Scale is subjective and unreliable.
• Objective quantification of PD tremor is essential for effective treatment.

Purpose of the Study:

• To develop a novel, objective system for measuring Parkinson's disease tremor severity.
• To evaluate the influence of different data collection tasks on tremor classification performance.
• To identify an optimal system comprising tasks, classifiers, and data preprocessing techniques.

Main Methods:

• A novel system was developed, evaluating six classifiers, six resampling techniques, and signal processing/feature extraction.
• The approach utilized an above-average rule based on five advanced metrics across four subdatasets.
• Task performance impact on classification was analyzed, comparing tasks with and without direct wrist movements.

Main Results:

• Tasks excluding direct wrist movements demonstrated superior performance for tremor severity measurement compared to wrist-focused tasks.
• Resampling techniques, particularly BorderlineSMOTE, significantly enhanced classification accuracy.
• An optimal system configuration was identified, including specific tasks and a Support Vector Machine (SVM) classifier.

Conclusions:

• The study recommends a system combining specific non-wrist-movement tasks (sitting, stairs, walking, standing) with an SVM classifier and BorderlineSMOTE for accurate PD tremor assessment.
• Objective tremor measurement systems are vital for improving PD management and treatment efficacy.
• Task selection during data collection critically influences the reliability of PD tremor quantification.